Nox Prompt Combiner

Nox Prompt Combiner is the pack's blunt instrument: two (or three) prompts in, one string out, with four combination modes. It's the node you grab when you have a character prompt from one source and a scene prompt from another and you just need them joined without hand-editing the commas. It's also the simplest node in the pack next to the Analyzer, and that's fine - sometimes you want a screwdriver, not a drill.

How it works

This is one of the few nodes here where the mechanism is genuinely worth understanding, because the four modes are not all created equal. Read the source and you get:

• concatenate (default) - joins the prompts with your separator (default , ). Boring, predictable, usually what you want.
• blend - interleaves the words of the two prompts one word at a time, position by position. It produces grammatically broken mash-ups, but for style blending that's sometimes the point: the model reads a hybrid of both vocabularies.
• alternate - alternates sentence by sentence between the prompts.
• weighted - this is the trap. It repeats the prompts' words proportional to weight_1 and weight_2 (each 0.0–2.0, default 1.0). The ComfyUI-familiar move is to think "weighted = emphasis," but that's not what this does - it literally duplicates words, so it inflates token counts. On modern LLM-encoded models there's no per-word weight syntax being produced here; you're just making the prompt longer. If you want real attention weighting, use actual (word:1.3) syntax or an SDXL-lineage model where that syntax works.

Inputs: prompt_1, prompt_2, combination_mode, weight_1, weight_2 are required; prompt_3 and separator are optional. One output: combined_prompt. That's the whole node. No presets, no notes, no safety rails - it's deliberately the dumb one.

Inputs and outputs that matter

For most workflows you want concatenate with , as the separator. Use alternate if you're splicing two narratively distinct paragraphs. Skip weighted unless you know you want token duplication, and treat blend as an experiment button.

Installing it

Pack install, once. ComfyUI Manager: search "Nox Prompter", install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesk9526/NoxPrompter-ComfyUI-CustomNode

Restart, find the NoxPrompter category. Zero dependencies.

Gotchas

The weighted mode misunderstanding above is the main one - don't expect per-prompt emphasis, because there's no weight syntax in the output, just repetition. Also worth noting: this node doesn't do contradiction detection or deduplication the way Nox Prompt Pipeline Combiner does, so if you combine two prompts that both say "daytime," you get "daytime" twice. If you're feeding a big assembled prompt into an LLM-encoded model, that's usually fine; if you're on a 77-token CLIP model, watch the token count - a long concatenation can silently push your subject past the token window. And as always with this pack, the README's stale ComfyNoxPrompter URL is not the repo you want; clone from the command above.